Wechar created HIVE-28200: ----------------------------- Summary: Improve get_partitions_by_filter/expr when partition limit enabled Key: HIVE-28200 URL: https://issues.apache.org/jira/browse/HIVE-28200 Project: Hive Issue Type: Improvement Components: Hive Reporter: Wechar Assignee: Wechar
When {{hive.metastore.limit.partition.request}} is configured, HMS would get the matching partition counts before get the real partition objects. The count could be a slow query if the input filter or expr is too complex. In this case, such slow filter will be executed in both counting partition numbers and fetching real partition objects, which harms the performance and backend DBMS. We can make an improvement by getting matched partition names firstly, and then check limit through the size of partition names, and finally get the partitions by the partition names. -- This message was sent by Atlassian Jira (v8.20.10#820010)